Announcements Monday, November 26

Transcription

1 Announcements Monday, November 26 Please fill out your CIOS survey! WeBWorK 6.6, 7.1, 7.2 are due on Wednesday. No quiz on Friday! But this is the only recitation on chapter 7. My office is Skiles 244 and Rabinoffice hours are: Mondays, 12 1pm; Wednesdays, 1 3pm.

2 Orthogonal Projections Review Recall: Let W be a subspace of R n. The orthogonal complement W is the set of vectors orthogonal to everything in W. The orthogonal decomposition of a vector x with respect to W is the unique way of writing x = x W + x W for x W in W and x W in W. The vector x W is the orthogonal projection of x onto W. It is the closest vector to x in W. To compute x W, write W as Col A and solve A T Av = A T x; then x W = Av. x W x W x W W

3 Projection as a Transformation Change in Perspective: let us consider orthogonal projection as a transformation. Definition Let W be a subspace of R n. Define a transformation T : R n R n by T (x) = x W. This transformation is also called orthogonal projection with respect to W. Theorem Let W be a subspace of R n and let T : R n R n be the orthogonal projection with respect to W. Then: 1. T is a linear transformation. 2. For every x in R n, T (x) is the closest vector to x in W. 3. For every x in W, we have T (x) = x. 4. For every x in W, we have T (x) = T T = T. 6. The range of T is W and the null space of T is W.

4 Projection Matrix Method 1 Let W be a subspace of R n and let T : R n R n be the orthogonal projection with respect to W. Since T is a linear transformation, it has a matrix. How do you compute it? The same as any other linear transformation: compute T (e 1), T (e 2),..., T (e n).

5 Projection Matrix Example Problem: Let L = Span {( 3 2)} and let T : R 2 R 2 be the orthogonal projection onto L. Compute the matrix A for T. It s easy to compute orthogonal projection onto a line: T (e 1) = (e 1) L = u e1 u u u = 3 ( ) T (e 2) = (e 2) L = u e2 u u u = 2 ( ) = A = 1 ( )

6 Projection Matrix Another Example Problem: Let x 1 W = x 2 in R 3 x1 x 2 + x 3 = 0 x 3 and let T : R 3 R 3 be orthogonal projection onto W. Compute the matrix B for T. In the slides for the last lecture we computed W = Col A for 1 1 A = To compute T (e i ) we have to solve the matrix equation A T Av = A T e i. We have ( ) ( ) A T 2 1 A = A T e 1 2 i = the ith column of A T =

7 Projection Matrix Another Example, Continued Problem: Let x 1 W = x 2 in R 3 x1 x 2 + x 3 = 0 x 3 and let T : R 3 R 3 be orthogonal projection onto W. Compute the matrix B for T. ( 2 1 ) 1 RREF ( 1 0 ) 1/ /3 ( ( ) RREF ( 1 0 ) 2/ /3 = T (e 1 ) = 1 ( ) 1 3 A = = T (e 2 ) = 1 ( ) 2 3 A = = T (e 2 ) = 1 ( ) 1 3 A = = B = ) RREF ( 1 0 ) 1/ /3

8 Projection Matrix Method 2 Theorem Let {v 1, v 2,..., v m} be a linearly independent set in R n, and let A = v 1 v 2 v m. Then the m m matrix A T A is invertible. Proof: We ll show Nul(A T A) = {0}. Suppose A T Av = 0. Then Av is in Nul(A T ) = Col(A). But Av is in Col(A) as well, so Av = 0, and hence v = 0 because the columns of A are linearly independent.

9 Projection Matrix Method 2 Theorem Let {v 1, v 2,..., v m} be a linearly independent set in R n, and let A = v 1 v 2 v m. Then the m m matrix A T A is invertible. Let W be a subspace of R n and let T : R n R n be the orthogonal projection with respect to W. Let {v 1, v 2,..., v m} be a basis for W and let A be the matrix with columns v 1, v 2,..., v m. To compute T (x) = x W you solve A T Av = Ax; then x W = Av. v = (A T A) 1 (A T x) = T (x) = Av = [ A(A T A) 1 A T ] x. If the columns of A are a basis for W then the matrix for T is A(A T A) 1 A T.

10 Projection Matrix Example Problem: Let L = Span {( 3 2)} and let T : R 2 R 2 be the orthogonal projection onto L. Compute the matrix A for T. The set {( 3 2)} is a basis for L, so A = u(u T u) 1 u T = 1 u u uut = 1 13 ( ) 3 ( ) = 2 13 ( ) Matrix of Projection onto a Line If L = Span{u} is a line in R n, then the matrix for projection onto L is 1 u u uut.

11 Projection Matrix Another Example Problem: Let x 1 W = x 2 in R 3 x1 x 2 + x 3 = 0 x 3 and let T : R 3 R 3 be orthogonal projection onto W. Compute the matrix B for T. In the slides for the last lecture we computed W = Col A for 1 1 A = The columns are linearly independent, so they form a basis for W. Hence ( ) 1 B = A(A T A) 1 A T 2 1 = A A T = 1 ( ) A A T 1 2 =

12 Poll Let W be a subspace of R n. Poll Let A be the matrix for proj W. What is/are the eigenvalue(s) of A? A. 0 B. 1 C. 1 D. 0, 1 E. 1, 1 F. 0, 1 G. 1, 0, 1 The 1-eigenspace is W. The 0-eigenspace is W. We have dim W + dim W = n, so that gives n linearly independent eigenvectors already. So the answer is D.

13 Projection Matrix Facts Theorem Let W be an m-dimensional subspace of R n, let T : R n W be the projection, and let A be the matrix for T. Then: 1. Col A = W, which is the 1-eigenspace. 2. Nul A = W, which is the 0-eigenspace. 3. A 2 = A. 4. A is similar to the diagonal matrix with m ones and n m zeros on the diagonal. Proof of 4: Let v 1, v 2,..., v m be a basis for W, and let v m+1, v m+2,..., v n be a basis for W. These are (linearly independent) eigenvectors with eigenvalues 1 and 0, respectively, and they form a basis for R n because there are n of them. Example: If W is a plane in R 3, then A is similar to projection onto the xy-plane:

14 A Projection Matrix is Diagonalizable Let W be an m-dimensional subspace of R n, let T : R n R n be the orthogonal projection onto W, and let A be the matrix for T. Here s how to diagonalize A: Find a basis {v 1, v 2,..., v m} for W. Find a basis {v m+1, v m+2,..., v n} for W. Then m ones, n m zeros A = v 1 v 2 v n v 1 v 2 v n Remark: If you already have a basis for W, then it s faster to compute A(A T A) 1 A T.

15 A Projection Matrix is Diagonalizable Example Problem: Let x 1 W = x 2 in R 3 x1 x 2 + x 3 = 0 x 3 and let T : R 3 R 3 be orthogonal projection onto W. Compute the matrix B for T. As we have seen several times, a basis for W is 1 1 1, By definition, W is the orthogonal complement of the line spanned by (1, 1, 1), so W = Span{(1, 1, 1)}. Hence B = =

16 Reflections Let W be a subspace of R n and let x be a vector in R n. Definition The reflection of x over W is the vector ref W (x) = x 2x W. In other words, to find ref W (x) one starts at x, then moves to x x W = x W, then continues in the same direction one more time, to end on the opposite side of W. Since x W = x x W we have x x W x W W x W ref W (x) ref W (x) = x 2(x x W ) = 2x W x. If T is the orthogonal projection, then ref W (x) = 2T (x) x.

17 Reflections Properties Theorem Let W be an m-dimensional subspace of R n, and let A be the matrix for ref W. Then 1. ref W ref W is the identity transformation and A 2 is the identity matrix. 2. ref W and A are invertible; they are their own inverses. 3. The 1-eigenspace of A is W and the 1-eigenspace of A is W. 4. A is similar to the diagonal matrix with m ones and n m negative ones on the diagonal. 5. If B is the matrix for the orthogonal projection onto W, then A = 2B I n. Example: Let The matrix for ref W A = x 1 W = x 2 in R 3 x 1 x 2 + x 3 = 0. x 3 is I 3 =

18 Summary Today we considered orthogonal projection as a transformation. Orthogonal projection is a linear transformation. We gave three methods to compute its matrix. Four if you count the special case when W is a line. The matrix for projection onto W has eigenvalues 1 and 0 with eigenspaces W and W. A projection matrix is diagonalizable. Reflection is 2 projection minus the identity.